Investigation into Huawei's Safe City program, which pairs cameras with analysis software in 50+ countries, shows Huawei has overstated drops in crime
Sheridan Prasso / Bloomberg : Tweets: @bw , @hillmanje , and @sheridanasia Tweets: @bw : Huawei's surveillance networks appear to be failing to slow the rise of violent crime rates, in contrast to the company's rosy assessment https://www.bloomberg.com/... Jon Hillman / @hillmanje : Huawei's “Safe City” technology promised to reduce crime in Islamabad. But $100m and 1,950 surveillance cameras later...murders, kidnappings, and burglaries are up. | by @SheridanAsia citing @CSIS @ReconAsia data https://www.bloomberg.com/... Sheridan Prasso / @sheridanasia : Huawei is selling its “Safe City” surveillance technology to countries in the developing world at great cost. Yet it seems that all those surveillance cameras don't actually make cities safer after all. Read my latest here: https://www.bloomberg.com/...
Context & Ripple Effects
Bloomberg's investigation lands against a decade of expansion: Huawei's Safe City program has grown from flagship deployments like Uganda's facial recognition debut into a footprint spanning 50+ countries and 200+ cities, sold on the promise that paired cameras and analysis software reduce crime. The Islamabad case is the sharpest counterexample — $100 million and 1,950 cameras later, murders, kidnappings, and burglaries are up, with CSIS data cited against the company's own rosy assessments.
The finding matters because the sales model itself is under scrutiny: reporting on free or discounted equipment offers across Latin America showed how aggressively Huawei and ZTE seeded demand among cost-constrained governments, while later coverage of Moscow's 217K-camera network reframed such systems as instruments of digital repression rather than crime control. An overstated efficacy claim strikes at the last defensible justification for these purchases.
First-order effects
- Cities that bought the pitch — Islamabad most concretely — are left holding expensive camera networks whose stated purpose the evidence contradicts, weakening the case for further municipal spending on Safe City contracts.
- Huawei's core sales claim, that its analysis software demonstrably cuts violent crime, now faces direct contradiction from cited third-party data, raising the burden of proof for every pending Safe City bid.
Second-order effects
- Competing surveillance vendors can no longer lean on the same unexamined crime-reduction statistics, forcing the whole category toward verifiable outcome metrics as a condition of government contracts.
- Governments that accepted subsidized hardware as a low-cost entry point face a repricing question: if the promised safety dividend doesn't materialize, the discounted equipment looks less like aid and more like an obligation.
Third-order effects
- If efficacy claims keep failing audit, the durable function of exported camera networks shifts openly from crime prevention to state monitoring — the trajectory already visible in Moscow — pushing procurement debates toward governance safeguards rather than vendor benchmarks.
- The pattern points toward independent evaluation becoming a standard gate for public-safety AI purchases, with legislatures and auditors, not vendors' marketing decks, deciding whether surveillance expansions proceed.
The trend: Exported AI surveillance is being re-judged from a crime-fighting product into a governance risk, with efficacy audits and repression concerns converging to slow the Safe City playbook worldwide.